This repository is built on,
Please, follow the intstructions carefully when installing EG3D.
Run the following commands:
cd dataset_preprocessing/ffhq
python preprocess_in_the_wild.py --indir=INPUT_IMAGE_FOLDER
The projector needs vgg16 for loss computation, you can download vgg16.pt from https://nvlabs-fi-cdn.nvidia.com/stylegan2-ada-pytorch/pretrained/metrics/vgg16.pt and save it to eg3d/network
.
Run the the following for first frame projection:
cd eg3d
python run_projector.py --outdir=projector_out --latent_space_type w --network=networks/ffhq512-128.pkl --sample_mult=2 --image_path ./projector_test_data/frame001.png --c_path ./projector_test_data/frame001.npy
Notice: before you run the PTI, please run the w to get the latent code of the first frame.
Then run:
cd eg3d/projector/PTI
python run_pti_video.py --in_dir PATH/TO/DIRECTORY --run_name Experiment_name
You can generate a video of the results using:
python saud_gen_video.py --in_dir PATH/TO/DIRECTORY